Abstract.- Field and laboratory 

 procedures were used to acquire dol- 

 phin school size estimates from ver- 

 tical aerial photographs. Multiple 

 photographs were taken of 48 sepa- 

 rate schools during a 1989 eastern 

 tropical Pacific (ETP) dolphin abun- 

 dance survey. During a 12-week 

 "counting period," three readers did 

 independent counts of dolphins in 

 the photographs. For each school, the 

 best photograph imagery was se- 

 lected and the mean of the three in- 

 dependent counts was used to esti- 

 mate its "true" size. The coefficient 

 of variation (CV) for school size esti- 

 mates (between-reader precision) av- 

 eraged 5.4% and ranged between 

 1.2% and 14.6%. Most (92%) of the 

 schools were estimated with preci- 

 sion, resulting in a CV of less than 

 9.0%. Within-reader CV averaged 

 3.5% and ranged 1.4%-7.1%, indi- 

 cating that readers were quite pre- 

 cise. To test if reader methods were 

 constant during the counting period, 

 temporal trends in estimates were 

 tested by linear regression analyses 

 and a repeated-counts experiment 

 with repeated measures analysis of 

 variance (RM-ANOVA). Regression 

 analyses indicated no significant 

 temporal trends or bias in the de- 

 viation of counts from the means. 

 The RM-ANOVA showed a signifi- 

 cant "reader with time" interaction 

 which was attributed to the rela- 

 tively high variability between read- 

 ers in counts made at the start of 

 the experiment. Results suggested 

 that methods were constant and 

 counts were precise after an initial 

 "warm-up" counting session. 



Method and precision \n estimation 

 of dolphin school size with vertical 

 aerial photography 



James W. Gilpatrick, Jr. 



Southwest Fisheries Science Center 

 National Marine Fisheries Service, NOAA 

 8604 La Jolla Shores Drive, La Jolla, CA 92038 



Manuscript accepted 11 May 1993. 

 Fishery Bulletin 91:641-648 ( 1993). 



Visual estimation of the number of 

 dolphins in a school is difficult be- 

 cause dolphins dive and the entire 

 school is rarely visible at the sea sur- 

 face at one time. Shipboard observer 

 estimates can be highly variable and 

 they may be biased (Scott et al., 1985; 

 Anganuzzi and Buckland, 1989). Con- 

 sequently, estimates of dolphin popu- 

 lation abundance derived from visual 

 survey data may be biased, and re- 

 sulting management decisions aimed 

 at conserving these populations may 

 be inappropriate. 



The Southwest Fisheries Science 

 Center (SWFSC) has been conduct- 

 ing surveys to monitor temporal 

 trends in the abundance of eastern 

 tropical Pacific (ETP) populations of 

 the pantropical spotted dolphin 

 iStenella attenuata), spinner dolphin 

 (S. longirostris), striped dolphin (S. 

 coeruleoalba) and common dolphin 

 (Delphinus delphis). These surveys 

 are one part of a multifaceted effort 

 to conserve ETP dolphins subjected 

 to incidental mortality in the inter- 

 national purse-seine fishery for yel- 

 lowfin tuna, Thunnus albacares 

 (Perrin, 1975; Wade and Gerrodette, 

 1992). Starting in 1987, in response 

 to the problem of potential bias in 

 visual school-size estimates, annual 

 ETP surveys were complemented 

 with aerial photography of dolphin 

 schools. This allowed shipboard- 

 observer estimates to be compared 

 with estimates (of the same schools) 

 taken from counts of dolphins in the 

 photographs. 



School size estimates derived from 

 large-format (126-mm) aerial photo- 



graphs were validated during a 1979 

 study when five separate schools (size 

 range; 161-396 dolphins) were pho- 

 tographed and then captured in a 

 tuna purse-seine net (Scott et al., 

 1985). Results showed that dolphin 

 counts from the photographs were 

 not statistically different from counts 

 tallied by hand as dolphins were be- 

 ing released from the net. This sug- 

 gested that counts from aerial pho- 

 tographs approximated "true" school 

 size. 



This report details photographic 

 and counting methods used to derive 

 dolphin 1 school size estimates from 

 large-format vertical aerial photo- 

 graphs. Many of the techniques origi- 

 nated with Scott et al. (1985). The 

 techniques were modified to support 

 the photography and laboratory ef- 

 forts associated with large-scale ETP 

 dolphin population surveys (i.e., an- 

 nual surveys of 120 days; covering a 

 19-million-km 2 study area). After a 

 survey in 1989, three readers did in- 

 dependent dolphin counts from ap- 

 proximately 200 dolphin school 

 photographs (multiple photographs 

 were taken for 48 separate schools) 

 over 12 consecutive weeks. For each 

 school, the best photograph imagery 

 was selected (according to criteria de- 

 scribed below) and the mean of the 

 three independent counts was used 

 to estimate its true size. The coeffi- 

 cient of variation (CV) was used to 

 characterize the precision (between- 



'For purposes of this paper, "dolphin" refers to 

 dolphins as well as small toothed-whales that 

 are included in the photograph sample. 



641 



